One API that extracts text, reads scans and validates the result. Runs entirely on our own servers: your documents stay yours.
No SDK and nothing to install on your side. Plain HTTP all the way.
Send a PDF or image to the API, or drop it into your dashboard.
Text extraction or OCR on our own machines, via a queue.
Fetch JSON, text or a searchable PDF, or receive a webhook.
Everything you can do with documents, through the screen and through the API.
API-key access, limits per plan, OpenAPI description.
Text and metadata from PDFs, per page, in reading order.
Turn scans and photos into text and a searchable PDF.
A callback to your system as soon as a document is done.
Download the result as a searchable PDF.
Detect tables and return them as rows and columns.
Invoice number, amounts, dates and supplier, with validation of the result.
Split, merge, rotate, delete pages, compress, watermark and convert to images.
Permanently black out personal data, addresses and bank details.
Full-text search across all your documents.
DocuDex recognizes known documents itself, fast and private. Only what it is unsure about can go to an AI. That teaches your system something new, and after your approval DocuDex handles it itself next time.
Run a model on your own server or network (Ollama, LM Studio and others). Nothing leaves your environment.
Local model, Claude or no AI. Off by default; the administrator enables it per user, and users can set up their own AI.
With Claude, DocuDex asks permission per document. Every AI result stays flagged for review.
From scans and photos to Office, email and e-invoices: it all goes through the same API.
PNGJPGGIFBMPTIFWEBPHEIC
DOCXXLSXPPTXODTODSODPEPUBRTF
TXTMDLOGCSVTSVJSON
HTMLXMLEML
XML is recognised as an e-invoice (UBL and CII), also when embedded in a PDF.
One key, one endpoint, a predictable result. Examples in your own language.
{
"id": "2420232f-cf3f-…",
"status": "done",
"pages": 3,
"characters": 214,
"pages_text": [
{ "page": 1,
"text": "Notanummer 1234567890 …" }
],
"warnings": []
}
Pick the plan that fits. Change or cancel at any time.
Paid plans are activated manually after your request for now.
Buy a bundle of credits and pay 1 credit per job. You get every feature of the Pro plan, no subscription.
If it goes wrong because of a fault on our side, you get your credits back right away. If a file truly cannot be recognized, or only produces unusable output, send it to us for inspection first. We look at it to improve our process and you get your credits back double.
On our own servers. By default no external OCR or AI service sits in between; only if you choose AI reading with Claude does the content of that document go to Anthropic (you confirm that per document). Files are deleted automatically after your plan’s retention period, and you can delete them yourself through the API at any time.
PDF, images (PNG, JPG, GIF, BMP, TIFF, WEBP, HEIC), Office (Word, Excel, PowerPoint and OpenDocument), EPUB, RTF, text files (TXT, MD, CSV, TSV, JSON, HTML, XML) and email (EML), including e-invoices (UBL/CII). The maximum file size depends on your plan.
Request the plan and we activate it after confirmation. Automatic payment follows later.
If it goes wrong because of a fault on our side, you get your credits back right away. If a file truly cannot be recognized, or only produces unusable output, send it to us for inspection using the button on that job. We look at it to improve our process and you get your credits back double. If the job succeeds but the result is different from what you hoped, no credits are returned.
The API answers with a clear error code (402 or 429). Your usage is always visible in your account and you can upgrade whenever you like.
Create a free account and use the API within a minute.
Start for free →